The Irrawaddy River Sediment Flux to the Indian Ocean: the Original Nineteenth-Century Data Revisited

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by St Andrews Research Repository

The Irrawaddy River Sediment Flux to the Indian Ocean: The Original Nineteenth-Century Data Revisited

R. A. J. Robinson, M. I. Bird, Nay Win Oo,1 T. B. Hoey,2 Maung Maung Aye,3 D. L. Higgitt,4 Lu X. X.,4 Aung Swe,5 Tin Tun,6 and Swe Lhaing Win3

School of Geography and Geosciences, University of St. Andrews, St. Andrews, Fife KY16 9AL, United Kingdom (e-mail: [email protected])

ABSTRACT The Irrawaddy (Ayeyarwady) River of Myanmar is ranked as having the fifth-largest suspended load and the fourth- highest total dissolved load of the world’s rivers, and the combined Irrawaddy and Salween (Thanlwin) system is regarded as contributing 20% of the total flux of material from the Himalayan-Tibetan orogen. The estimates for the Irrawaddy are taken from published quotations of a nineteenth-century data set, and there are no available published data for the Myanmar reaches of the Salween. Apart from our own field studies in 2005 and 2006, no recent research documenting the sediment load of these important large rivers has been conducted, although their contribution to biogeochemical cycles and ocean geochemistry is clearly significant. We present a reanalysis of the Irrawaddy data from the original 550-page report of Gordon covering 10 yr of discharge (1869–1879) and 1 yr of sediment concentration measurements (1877–1878). We describe Gordon’s methodologies, evaluate his measurements and calculations and the adjustments he made to his data set, and present our revised interpretation of nineteenth-century discharge and sediment load with an estimate of uncertainty. The 10-yr average of annual suspended sediment load currently cited in the literature is assessed as being underestimated by 27% on the basis of our sediment rating curve of the nineteenth- century data. On the basis of our sampling of suspended load, the nineteenth-century concentrations are interpreted to be missing about 18% of their total mass, which is the proportion of sediment recovered by a 0.45-mm filter. The MT. Our revised estimate of the annual sediment load 60 ע new annual Irrawaddy suspended sediment load is364 from the Irrawaddy-Salween system for the nineteenth century (600 MT) represents more than half the present-day Ganges-Brahmaputra flux to the Indian Ocean. Since major Chinese rivers have reduced their load due to damming, the Irrawaddy is likely the third-largest contributor of sediment load in the world.

Introduction The eastern syntaxis of the Himalayas and the Ti- ween, Irrawaddy, Ganges, and Brahmaputra rivers, betan Plateau contains the most tectonically com- and two of these systems, the Ganges-Brahmaputra plex geology in Asia (Socquet and Pubellier 2005). (G-B) and the Irrawaddy-Salween (I-S), debouch into This region is drained by the Red, Mekong, Sal- the Indian Ocean. We consider the Irrawaddy and Salween with the Sittoung and four smaller rivers Manuscript received January 26, 2007; accepted June 18, together because their catchments are adjacent and 2007. both ﬂow into the Gulf of Martaban (ﬁg. 1) along 1 Department of Geography, University of Pyay, Pyay, a similar length of coastline to the G-B river sys- Myanmar. tem. The headwaters of the I-S are undergoing some 2 Department of Geographical and Earth Sciences, University of Glasgow, Glasgow G12 8QQ, United Kingdom. of the highest rates of landscape adjustment in a 3 Department of Geography, Yangon University of Distance region of active orogenesis (Zeitler et al. 2001). Education, Yangon, Myanmar. Collision, uplift, and shear zone development 4 Department of Geography, National University of Singa- since the early Tertiary have resulted in a complex pore, Kent Ridge 117570, Singapore. pattern of river capture involving the Irrawaddy, 5 Department of Geography, University of Yangon, Yangon, Myanmar. Salween, Red, Yangtze, Mekong, and Brahmaputra 6 Department of Geography, University of Mawlamyine, rivers (Clark et al. 2004). These large rivers deliver Mawlamyine, Myanmar. material from the Himalayas to the ocean, and

629 630 R. A. J. ROBINSON ET AL.

close to half of the current total flux of water, sediment, and dissolved load from the Himalayas and Tibet to the ocean, with ∼20% of this attributed to the I-S (Milliman and Meade 1983). The Irrawaddy River ranks fifth in the world in terms of suspended load (265 MT yrϪ1) and fourth in terms of dissolved load (Milliman and Meade 1983, their table 2); these published sediment and water fluxes are derived from a nineteenth-century data set (Gordon 1885), and the dissolved loads were published in an article by Meybeck and Ragu (1995). Stamp (1940) states that 48 MT yrϪ1 should be added as dissolved load for the Irrawaddy but gives no source for this estimate. Values published by Meybeck and Ragu (1995), Meade (1996), Gaillardet et al. (1999), and Meybeck et al. (2003) equate to combined I-S annual averages of 697 km3 of water, 365 MT of suspended material, and 162 MT of dissolved material. Although figures for these rivers are widely quoted and used for global flux calculations of carbon (e.g., Subramanian and Ittekkot 1991), there are no measurements of suspended load for the Salween in Myanmar, and the values quoted by Meade (1996, his table 1) are estimates based on the measured suspended load values of the Irrawaddy and Me- kong. The increase in damming along the Yangtze (Changjiang) and Yellow (Huanghe) rivers has sub- stantially reduced the sediment load of these rivers (Walling 2006), such that the Irrawaddy may now rank as having as high as the third-largest sediment load in the world after the Amazon and the G-B. We have reanalyzed the original nineteenth- century Irrawaddy data and subsequent early twentieth-century engineering reports that identify problems with the data collection. In their seminal article, Milliman and Meade (1983) use the Irra- waddy as an example of the potential error that can arise when compiling global sediment budgets us- Figure 1. Location of the original Irrawaddy study site ing data recycled from previous reports rather than of Gordon (1879–1880, 1885) at Seiktha and the rivers original, often inaccessible, data. Given the very included in this article. Our measurement sites for mod- significant contribution of the Irrawaddy River to ern discharge and material fluxes are at Seiktha and Pyay global sediment budgets, we have begun to compare on the Irrawaddy and at Hpa-An on the Salween. the revised nineteenth-century water and sediment load values to our own field measurements col- knowledge of the magnitude of fluxes, as well as lected during the onset of the 2005 and 2006 mon- the mineralogical and chemical characteristics of soon seasons. We use our preliminary data set here the transported material, is a prerequisite for un- simply to compare the efficiency of Gordon’s and derstanding the impact of Himalayan uplift on our filtering methods; we do not present any com- global biogeochemical cycles (Raymo et al. 1988) parison of nineteenth-century and modern sedi- and the impact of human activities on recent-mod- ment load values because this is the focus of on- ern sediment yields (Wilkinson 2005). An undergoing research. standing of denudation rates in areas of active tec- tonism is critical in assessing the extent to which Study Area and Methods climate-driven erosion influences large-scale tectonic behavior (An et al. 2001; Zeitler et al. 2001). The Irrawaddy and Salween rivers are ∼2000 and Together, the G-B and I-S are thought to deliver ∼2800 km long and have drainage areas of Journal of Geology IRRAWADDY SEDIMENT FLUX 631

0.413 # 1066and 0.272 # 10 km2, respectively (Nyi Goldsmith Forshey at hydrological stations on the 1967; Bender 1983). The Irrawaddy catchment rock Mississippi River between 1848 and 1855 during types include Cretaceous to mid-Cenozoic flysch the U.S. government’s Mississippi Delta Survey. of the western Indo-Burman ranges, Eocene- Gordon made daily measurements of flow velocity Miocene and Quaternary sediments of the Myan- and depth and recorded stage during 1872–1873 and mar Central Basin, and the Late Precambrian and again in 1875, missing only Sundays, holidays, and Cretaceous-Eocene metamorphic, basic, and ultra- days of very bad weather. Channel width at Seiktha basic rocks of the eastern syntaxis of the Himala- ranges from ca. 700 m at low flow to ca. 1600 m yas. The Salween and eastern tributaries of the Ir- at peak flow. Flow depth was measured along a line rawaddy drain Precambrian, Oligocene-Tertiary normal to the main flow from a boat using a lead sedimentary, and acidic and metamorphic rocks of and line, and the boat position was surveyed by the eastern Shan Plateau (Bender 1983; Curray triangulation from the bank with theodolites. Flow 2005; Najman 2005; Socquet and Pubellier 2005). velocity was estimated from the time taken for Gordon (1885) presented monthly discharge data floats to travel between two cross-channel base- collected between 1869 and 1879 and 1 yr of sed- lines spaced about 60 m apart, and theodolites po- iment load data (1877–1878) to the Royal Geo- sitioned at the end of each baseline were used to graphical Society (RGS) without details of his field track float trajectories. The starting and final po- methods or sampling locations. Gordon’s original sitions of the floats were drafted as scaled drawings, report, containing his full daily discharge, rainfall, and the downstream component of velocity was and sediment concentration data set, survey maps, resolved trigonometrically from the time of travel. channel cross sections, and a detailed description Surface floats, suspended up to ∼1 m below the of sampling techniques (Gordon 1879–1880), was surface to avoid wind and waves, were used ini- located in the archives of the RGS in 2005; another tially. Gordon recognized the need for vertical ve- copy is located in the British Library in London. locity profiles and so sank floats on increasing These data have been digitally reconstructed using lengths of line in ∼1-m intervals down through the OCR software, and every number has been checked water column; the depth of the river at Seiktha against the original report. The data are of remark- never exceeded 25 m (Gordon 1879–1880). The ve- able quality, particularly because there is a full locity measurements were usually reported as the range of flows, including monsoon peak discharge. average of two or three readings. The float (called Sediment concentration was measured at three a double float) was a solid cylinder of wood (1 ft depths within the water column, including the [30 cm] in height and 6 in [15 cm] in diameter) with lower part of the water column. Milliman and a4 # 410 -in (# 10 -cm) hole on the outside into Meade (1983, p. 2) note that inadequate sampling which was inserted the amount of wet clay ballast at peak flows and from throughout the water col- needed to sink the cylinder to the required depth umn are among the most important sources of error (Gordon 1879–1880). Floats were suspended from a in sediment load data. thin cord (1/16 in [ca. 1.5 mm] diameter) attached Gordon, a civil engineer in charge of river works to a smaller wooden cylinder (6 in [15 cm] diameter in Myanmar during the late nineteenth century, and 1 in [25 mm] thick) that floated just below the investigated the magnitude and duration of flood water surface. Gordon grouped similar surface ve- events on the Irrawaddy for the government of In- locity values together to divide the channel cross dia, in order to calculate the required size and section into 10 subsections (the sections were al- height of flood embankments. Gordon selected tered slightly for overbank floods) and calculated Seiktha (fig. 1) as the main measurement site for the area of each subsection. He calculated the ar- velocity and sediment concentration because it is ithmetic mean velocity for each vertical profile the farthest downstream site above the delta that across the channel and mean surface velocity for is constrained by bedrock highs on both its western each subsection. Some subsections had more than and its eastern banks. Seiktha is 16 mi upstream one vertical velocity profile (and some appear to of Myanaung (fig. 1), where daily stage was recorded have none), and for days of continuous stage height, from 1869 to 1879 but where the width of the river he calculated the mean of the average vertical ve- made multiple cross-sectional measurements of locities for each profile. Gordon (1879–1880) notes sediment load, velocity, and depth too difficult. that his measurements at any location remained Nineteenth-Century Cross-Sectional Area and Flow the same on consecutive days with constant stage Velocity Measurements. Gordon’s (1879–1880) (and differed with stage) and varied both within a hydrological sampling strategy was based on the subsection if there were multiple profiles and flotation methods published in an article by across the whole cross section. About 10,000 ver- Humphreys and Abbot (1861), as used by Caleb tical profiles of float velocity were recorded by the 632 R. A. J. ROBINSON ET AL.

Figure 2. Annual hydrographs at Seiktha, using Gordon’s original (1879–1880) data. Daily discharge values for 1869– 1874 (A) and 1875–1879 (B). The timing, duration, and peak value of the monsoon seasons produce total annual .(from the mean (inset in A 11%ע discharges that vary by end of 1875, and each channel cross section (1 d of falling flows at Seiktha (fig. 2); he presented both measurements) was covered by about 10 profiles the measured and the predicted discharge values (Gordon 1879–1880). He then calculated discharge and states that in only very few cases do the dif- by summing the velocity-area product of each sub- ferences exceed 10% (Gordon 1879–1880, p. 22). section. He used 2 yr of daily data (excluding Sun- In addition to daily sampling at Seiktha for the days and holidays) and the stage readings to develop periods mentioned above, Gordon used a simple stage-discharge rating relationships for rising and transfer function to match the Myanaung stage rec- Journal of Geology IRRAWADDY SEDIMENT FLUX 633 ord (fig. 1) to that at Seiktha and then calculated stage record. Gordon (1879–1880, p. 22) compared daily discharge for the remaining years between the Seiktha and Myanaung records for 1872–1873 1869 and 1879 at Seiktha from the Myanaung stage and stated that the Myanaung record was “with measurements (fig. 2). Gordon used the Myanaung some uniformity about 1/2 foot lower” than at record instead of the Seiktha stage record because Seiktha and the difference did not vary over a range a sand bar developed downstream of Seiktha be- of flows. tween 1873 and 1875, and it raised the water level Gordon (1879–1880) discussed possible errors at the gauging location “by 2–3 feet” (Gordon 1879– with the double float method, including drag as- 1880, p. 22) and because Myanaung had the longer sociated with the cord, the accuracy of calculated

Figure 3. Original discharge and suspended sediment concentrations for 1877–1878. A, Concentrations are based on an average of surface, middle, and near-bottom depth suspended sediment sample measurements. B, Individual concentrations for each water depth; inset shows the ratio of the average of surface and middle sediment concentrations to bottom concentration. 634 R. A. J. ROBINSON ET AL.

time over the short distance, and eye fatigue of the observers. In 1882, Gordon cross-checked the double float method with an electrical Deacon current meter and found that the water velocities from lower parts of the water column were less than his first float measurements and that the difference increased with flood level. He consequently sug- gested reducing his calculated discharges (Gordon 1879–1880) by 10%, 5%, and 0% for high, medium, and low discharge ranges, respectively, and applied these corrections to the discharge data set presented to the RGS in 1885. He did not discuss the numerical values of high, medium, and low discharges. Subsequent engineers reevaluated Gordon’s data, particularly Samuelson (1913), who took over the Embankment Division in 1913. Samuelson noted several problems with Gordon’s data, in particular that the Myanaung and Seiktha stage records varied with changing channel geometry at Myanaung, producing systematic errors in the transfer function between the two locations. One of the features of Gordon’s data set (fig. 2) is an apparent increase in discharge for the years after 1875 when the stage records were merged. We discuss this source of uncertainty in “Review of Potential Sources of Error.” Nineteenth-Century Sediment Load Measurements. Gordon measured sediment load for 6 d each week (excluding Sundays, holidays, and storms) for the year of 1877–1878 (52 wk) at three positions across the river: ∼90 m from the right bank, in midstream, and ∼1220 m from the right bank. At the first two positions, samples were taken at three depths: surface (∼1 m depth), middle, and near-bottom (0.5–1 m above the riverbed). The left bank position had only surface and lower measurements as a result of the shallow water depths there. He used a sediment sampler that was based on the device de- signed for the Mississippi Delta Survey (Hum- phreys and Abbot 1861). Gordon collected water using a hollow barrel (1 ft [30 cm] long and 6 in [15 cm] diameter) loaded with lead rings to make it sink. The barrel had two large valves on the top and bottom, and its lids were made of leather and lead, and they freely opened upward; as the barrel descended through the water column, the lids re-

Figure 4. A, Relationship between sediment concen- with our 2006 data set. C, Relationship between total tration in surface (CS) and middle depth (CM) levels of the mass measured in 2005 and 2006 and settled mass before water column for the 1877–1878 data, shown with our sieving with 10.45-mm nucleopore mesh. Average under- data from 2005 and 2006. B, Average of CS and CM (CSM) estimate of the historical Irrawaddy suspended concen- compared with near-bottom concentration (CB), shown tration values inferred from this data is 18%. Journal of Geology IRRAWADDY SEDIMENT FLUX 635 mained open, and water entered and passed through is 340 MT yrϪ1. These sediment load values (Gor- it, but the lids closed once the barrel was being don 1879–1880, his table 16) are based on the av- lifted out of the water. The sampler thus collected erage sediment concentrations from all water water from the lowest depth to which the barrel depths (fig. 3). Gordon (1879–1880) originally re- descended. Gordon combined 100 g of water and ported that the 10-yr average sediment load for sediment from every bottle collected over 6 d for 1869–1879 was 286 MT yrϪ1; this is close to the each of the three water depths to produce 52 weekly number quoted in table 1 of Milliman and Meade measurements of sediment concentration for each (1983). However, as discussed earlier, by 1885, Gor- depth. He filtered the combined samples (600 g of don had reduced the discharge measurements for water) through a double thickness of weighed filter the monsoon months, and the most frequently paper, and although he does not mention the qual- quoted annual sediment load value of 261 MT yrϪ1 ity or grade of paper, it seems unlikely that it was (i.e., Stamp 1940; Milliman and Syvitski 1992) a very small pore size because filtering such large comes from the 1885 article presented to the RGS volumes of water and sediment would have been (table 1 in Gordon 1885). extremely slow. The dried filter papers and sedi- Modern Discharge and Sediment Load Measure- ment were weighed carefully using a Fortin bal- ments. In June 2005, data were collected at Gor- ance; the filter paper alone was reweighed, com- don’s original site at Seiktha, including 10 bathy- pared to its original weight, and the sediment mass metric cross-profiles, 10 vertical velocity profiles was calculated. Only the combined weekly con- (measured using a Valeport velocimeter), and 18 centrations for each depth were reported by Gordon suspended sediment load samples collected using (1879–1880). Measurements were reported in grams a horizontally deployed, cylindrical 2-L Van Dorn of sediment per 100 g of water (1 g per 100 g being water sampler with openings at each end. We used equivalent to 104 mg LϪ1) for each depth and week a sediment sampling protocol similar to that of of measurement. The suspended sediment concen- Gordon (1879), involving collection of water and tration and discharge data are presented in figure 3 sediment from surface and middle depths (but not and will be discussed in “Reanalysis of Gordon’s bottom depth) and then settling sediment from 2 Original Data.” Gordon developed a concentration- L of water for 24 h. We decanted the water and discharge relationship for the 1877–1878 data but passed it through a 0.45-mm nucleopore filter to did not discuss how it was derived. This relation- determine the likely recovery efficiency of Gor- ship was combined with the 1869–1879 discharge don’s protocol. Separate quantification of the set- estimates to calculate annual sediment loads that tled and filtered components suggests that 18% of range from 248 to 352 MT yrϪ1; the directly mea- the sediment load remains suspended after 24 h of sured load for the year of measurement (1877–1878) settling and is recovered with a 0.45-mm filter (fig.

Table 1. Drainage Areas, Water Discharge, and Suspended Sediment Loads for Rivers Entering the Gulf of Martaban, Myanmar Suspended sediment flux (#106 TyrϪ1) Drainage Average surface Average surface, Drainage area length Water flux and middle middle, and (#106 km2) (km) (km3 yrϪ1) depth bottom depth Milliman and Meade’s (1983) data for I .43 … 428 … 265 (261a) … …(1j) 48 ע Reanalysis of I data from Gordon (1879–1880) .413b 1985 440 49 ע 309 32 ע 1j) 191) 41 ע Corrected data after velocity reductionc … … 422 60 ע 364 39 ע Addition due to unmeasured fine fraction (18%) … … … 226 Salween (Thanlwin) .272b 2800 211b 110c 180c Sittoung, Bago, Bilin, Attaran, and Gyaning .06b … 119b 30c 50c Total for I-S system .745 … 752 370 600 G-B system 1.59a 2700 971 1060a 1060a Percentage of estimated I-S relative to G-B (%) 47 … 77 35 57 Note.I p Irrawaddy ; I-S p Irrawaddy-Salween ; G-B p Ganges-Brahmaputra . a From Milliman and Syvitski (1992). Note that the sediment flux is used for the G-B in calculating the ratios of sediment flux for these systems because this is the only value available. b Drainage areas, river lengths, and Salween and Sittoung and smaller river discharge values from Meybeck and Ragu (1995), Bender (1983), and Nyi (1967). c The sediment loads of the Salween, Sittoung, and smaller rivers are estimated proportionally using discharge scaling and the Irrawaddy sediment load values.

Journal of Geology IRRAWADDY SEDIMENT FLUX 637

4). When we used average velocities and bathym- low 10,000 m3 sϪ1, “medium flows” are between etry for each of our profiles, the velocity-area 10,000 and 35,000 m3 sϪ1, and “high flows” are method gives a discharge of 5058 m3 sϪ1 on June greater than 35,000 m3 sϪ1. This reduces our cal- 26, 2005, carrying 252 mg LϪ1 of suspended sedi- culation of mean annual water flux by ∼4%, from km3 yrϪ1 (table 1); Gordon’s 41 ע 48to 422 ע ments with a median grain size of 10.5 mm. 440 In May 2006, we remeasured discharge with an 1885 reduced discharge value is 400 km3 yrϪ1, acoustic Doppler current profiler and collected wa- which represents an ∼6% reduction. We intend to ter and sediment samples using the same protocol quantify the uncertainty that should be attributed as in 2005, including sampling near-bottom water to Gordon’s double float method in future field sea- depths (0.5–1 m above the riverbed) from Seiktha sons in order to objectively correct the original and Pyay (40 mi upstream) and at Hpa-An on the nineteenth-century discharge values. Salween. Measurements of water discharge (and The daily concentration measurements for 1877– sediment concentration) made between May 23 and 1878 are presented as an average of surface, middle, Ϫ1 ע Ϫ1 3 May 30, 2006, were 5370 m s (mgL271 16 ) and near-bottom depth concentrations (CSMB; fig. Ϫ1 ע Ϫ1 3 at Seiktha, 5030 m s (mgL210 9 ) at Pyay, 3A); as individual surface (CS), middle depth (CM), Ϫ1 ע Ϫ1 3 and 2270 m s (mgL127 11 ) at Hpa-An. Our and near-bottom depth measurements (CB; fig. 3B); total sediment concentrations (settled and filtered and as ratios of CSM to CB (fig. 3B). Figure 4 dem- fractions) for June 2005 and May 2006 are plotted ∼ onstrates that the values of CS are 80% of those together with Gordon’s data in figure 4. This article measured at CM and that the values of CB are be- makes no statements about possible changes in sed- tween two and four times higher than those in the iment load between the nineteenth century and to- upper parts of the water column and display more day because we have insufficient data on which variability. Gordon suspected that some of his mea- such comparisons can be based. We do, however, surements of CB were overestimated because of bed discuss how sampling and filtering methods can disturbance. Our small data set collected in May affect measured sediment loads in “Review of Po- 2006 during the onset of monsoon (days 144–151) tential Sources of Error.” has ratios ofCSM/C B in the range of 0.86–1.1, which

may suggest that Gordon’s CB measurements are too high (fig. 4A). However, Gordon’s data set also Reanalysis of Gordon’s Original Data shows that the ratio ofCSM/C B approaches 1 during We have recalculated discharge (fig. 2) and sus- the onset of monsoon (May–June), which is when pended sediment load from the data in Gordon’s we sampled (fig. 3B). In the absence of a modern 1879–1880 report (fig. 3; table 1). Our 10-yr average sediment concentration data set that covers the full range of flows, which would allow us to evaluate ע of total annual discharge is calculated as 440 48 km3 yrϪ1 (table 1), which is higher than Gordon’s whether Gordon overestimated near-bottom sus- 3 Ϫ1 (1879–1880) value of 425 km yr , as a result of pended concentrations, CSM will be used as a lower arithmetic errors in his calculations. However, limit for the mean sediment concentration and

Gordon’s (1885) later publication includes a reduc- CSMB will be used as an upper limit. It is worthwhile tion of the original discharge measurements. Gor- repeating here that Gordon’s (1885) quoted sedi- don does not discuss exactly how he adjusted his ment load values are the average of the three depth discharge measurements, and the data are not com- concentration measurements. plete enough to allow a correction factor to be cal- We have calculated new sediment rating curves culated. He states that high, medium, and low flow for the adjusted discharge and sediment concentra- measurements were reduced by 10%, 5%, and 0%, tion values for 1877–1878 (fig. 5) based on a re- respectively, and by comparing the 1879 and 1885 gression of the log-log plot of discharge and sedi- publications, we estimate that “low flows” are be- ment concentration and including a bias correction

Figure 5. Sediment rating curves for 1877–1878 measurements. A, Regression of data derived from log-log plot using average of surface, middle, and bottom depths (CSMB). B, Linear plot showing interpretation of positive hysteresis for early monsoon period and negative hysteresis for late monsoon period. C, Cumulative annual and seasonal rating curves based on concentrations averaged over surface and middle depths (CSM) and CSMB compared with Gordon’s measured values. Modeled rating curve values of sediment concentration deviate from measured values for the late monsoon period. 638 R. A. J. ROBINSON ET AL. factor to account for the inherent underestimation surement technique, (2) the missing 10.45-mm fil- in total load that comes from using a log-log data tered fraction in the sediment concentration data, transformation; this implicitly makes allowance (3) the merging of two stage-discharge relationships for the poor fit of the regression to sediment con- developed at two different locations, and (4) the centrations at low discharges (Ferguson 1986). The contribution of CB to the daily depth-averaged con- sediment-discharge relationships are different for centration values. We have adjusted the raw dis- early and late monsoon periods; the former exhibits charge values to account for the uncertainty due to “first flush” positive hysteresis behavior with a the velocity measurement technique (an ∼4% re- steeper rating curve, while negative hysteresis is duction in the 10-yr average of total annual dis- displayed for the late monsoon period (fig. 5B). We charge) but recognize that Gordon’s method of mea- compared sediment concentrations calculated us- suring velocity needs to be independently verified. ing one annual rating curve (fig. 5A) with those Sediment concentrations have been increased calculated using two seasonal curves (fig. 5C). De- based on our estimation of the missing particulate spite the clear visual differences in the sediment fraction (an 18% increase). Our analysis of the stage concentration-discharge behavior between early records from Seiktha and Myanaung suggests that and late monsoon periods, a seasonal (two-curve) the merging of two stage records from different lo- relationship overpredicts sediment concentrations cations introduces an uncertainty of about 8%. Fi- for both CSM and CSMB, compared with the measured nally, the largest uncertainty in the data set is the values (fig. 5C). We have therefore used the annual values for CB. It is possible that some of the CB curve to calculate sediment concentrations for the measured by Gordon overrepresents the true bot- remaining 9 yr of the discharge records. Notably, tom suspended sediment concentration. The barrel our single annual sediment rating curve predicts a used in Gordon’s sampling probably had a larger sediment load of 337 MT for 1877–1878, which is internal diameter than the openings that let the the same as the measured value of 340 MT (fig. 5C) flow move through: therefore, as the barrel de- reported by Gordon (1879–1880). scended through the column, sand-sized particles Our filtered sediment concentration data from in suspension may have been trapped inside, thus 2005 and 2006 suggest that the concentrations cal- resulting in oversampling of the suspended sand- culated by Gordon did not include the 10.45-mm sized material. The amount of this oversampling particulate fraction. We therefore use an increase would increase with depth because the barrel col- of 18%, derived from our data set, as an estimated lects sand for a longer period of time in deeper flows increase to sediment load (a filtered correction). and because the sand concentration increases to-

The raw and adjusted Gordon data, using both CSM ward the bed. Another possibility is that during and CSMB, are presented in table 1. With no correc- sampling, the barrel could have disturbed the bed, tion for the missing particulate fraction, the Irra- producing bottom concentration values that are not -km3 yrϪ1 of water and entirely due to suspended load. We intend to ad 41 ע waddy transported422 -MT yrϪ1 of sediment load (10-yr averages). dress all these sampling issues in future field cam 49 ע 309 With our 18% increase for the missing fine partic- paigns by comparing various devices for measuring suspended load concentrations, including a barrel ע ulate fraction, the sediment load becomes 364 60 MT yrϪ1. with a design similar to that of Gordon’s (i.e., that In summary, the recalculated sediment load data of Humphreys and Abbot [1861]) and the Van Dorn differ from Gordon’s (1885) values because (1) there sampler. were minor arithmetic errors in the original nineteenth-century discharge calculations (accounting Implications for an ∼4% decrease in discharge); (2) our sediment rating curve produces a 27% increase in the 10-yr In order to estimate the suspended sediment load average annual sediment load measurement, even debouched into the Gulf of Martaban, we have though our rating curve correctly estimates the combined the new Irrawaddy load values with es- measured value for 1877–1878; and (3) our filtered timates for the Salween and Sittoung, as well as correction that accounts for a missing particulate four smaller rivers (Bilin, Bago, Attaran, and fraction in Gordon’s concentration measurements Gyaing) that terminate in the gulf (fig. 1). Their increases the loads by 18% (table 1). estimates are based on a scaling of the Irrawaddy annual sediment load values in proportion to their annual discharge relative to the Irrawaddy, and we Review of Potential Sources of Error recognize that this is a crude estimate in the ab- There are four sources of potential uncertainty in sence of more data; we therefore present these val- Gordon’s Irrawaddy data set: (1) the velocity mea- ues, and the total combined flux, as rounded to the Journal of Geology IRRAWADDY SEDIMENT FLUX 639 nearest two significant digits (table 1). Our lower in southeast Asia at the present time is absolutely value of suspended sediment load for the Salween critical. is 110 MT yrϪ1, similar to the 100–MT yrϪ1 estimate of Meade (1996), which was based on scaling values Conclusions for the Irrawaddy and Mekong. Our upper value of 180 MT yrϪ1 is similar to an annual suspended load The existing sediment flux data for the Irrawaddy of 164 MT, calculated from the yield value pre- River, one of six major systems draining the eastern sented by Meybeck et al. (2003) for the Salween in Himalaya, come from a comprehensive nineteenth- Thailand. The total annual suspended sediment century hydrological study, but these original data load to the eastern Indian Ocean for the Irrawaddy- have not been reevaluated since recognition of the Salween-Sittoung system is potentially as large as significance of these river systems to global climate 600 MT, representing 57% of the modern G-B sed- change and the evolution of the Himalayan orogen. iment flux (table 1); this represents sediment loads To date, only recycled values from early twentieth- calculated from three depths (average of CS, CM, and century summaries of the work have been incor-

CB) for the Irrawaddy. In comparison, because of the porated in global river sediment load and discharge uncertainty surrounding the near-bottom concen- databases. We have evaluated the sampling strat- tration measurements, total annual sediment load egies, reanalyzed the original data, and recalculated for the Irrawaddy based on averaging just the upper discharges and sediment loads and find that the and middle depths is 370 MT and represents 35% typically quoted sediment flux values (e.g., Milli- of the G-B flux. man and Meade 1983; Meade 1996) are underesti- Over the past century or more, many rivers mated. worldwide have undergone substantial declines in We therefore believe that the significance of the sediment export as a result of damming and irri- I-S contribution to “natural” Himalayan denuda- gation, but the results of Syvitski et al. (2005) sug- tion and land-ocean fluxes has not been fully ap- gest that the Irrawaddy has undergone the largest preciated because (1) reanalysis of the original nine- relative increase in sediment delivery to the ocean teenth-century data (Gordon 1879–1880) for the Irrawaddy suggests that the often-quoted flux of of any major river as a result of land use and pop- Ϫ1 ulation changes and the absence of dams on its 261 MT yr may be a significant underestimate, Ϫ1 ע mainstem or major tributaries. The majority of the with the true value as high as364 60 MT yr , which would place the Irrawaddy as third-largest Irrawaddy load was thought by Stamp (1940) to contributor of sediment load after the Amazon and have been derived from below Mandalay from the G-B (due to reduced loads on the major dammed unconsolidated and rapidly eroding “dry zone” Chinese rivers), and (2) the I-S system may con- (!800 mm annual rainfall) of the Central Basin of Ϫ tribute 35%–57% (370–600 MT yr 1) of the G-B Myanmar (fig. 1), where recent population and ag- load. ricultural changes have produced significant land use changes. We are currently developing modern mass flux records for the Irrawaddy and Salween, ACKNOWLEDGMENTS as well as methods that allow us to hindcast sed- We would like to thank the Royal Geographical iment loads back to the nineteenth century, in or- Society and the Carnegie Trust for the Universities der to address how land use and climate have in- of Scotland for funding this research, including the fluenced sediment loads. The sparsely inhabited purchase and installation of logging equipment and nature of the Salween catchment means that land the costs of fieldwork. We are very grateful for the use change has had less of an impact on this river, careful and thorough review of Bob Meade and to but an extensive damming and hydroelectrical John Milliman and an anonymous reviewer, whose scheme is scheduled to begin in 2007 (Mizzima comments also helped to improve the article. This News 2007), which will have a significant effect on publication is a contribution from the Scottish Al- future sediment flux. Measuring the sediment load liance for Geoscience, Environment, and Society of the, as yet, longest nondammed river remaining (http://www.sages.ac.uk).

REFERENCES CITED

An, Z.; Kutzbach, J. E.; Prell, W. L.; and Porter, S. C. 2001. Bender, F. 1983. Geology of Burma. Berlin, Borntraeger, Evolution of Asian monsoons and phased uplift of the 293 p. Himalaya-Tibetan plateau since Late Miocene times. Clark, M. K.; Schoenbolm, L. M.; Royden, L. H.; Whipple, Nature 411:62–66. K. X.; Burchﬁeld, B. C.; Zhang, X.; Wang, E.; and Chen, 640 R. A. J. ROBINSON ET AL.

L. 2004. Surface uplift, tectonics, and erosion of east- the importance of small mountainous rivers. J. Geol. ern Tibet from large-scale drainage patterns. Tectonics 100:525–544. 23:TC1006, doi:10.1029/2002TC001402. Mizzima News. 2007. EGAT resumes Hat Gyi Dam stud- Curray, J. R. 2005. Tectonics and history of the Andaman ies. http://www.mizzima.com. Sea region. J. Asian Earth Sci. 25:187–232. Najman, Y. 2005. The detrital record of orogenesis: a re- Ferguson, R. I. 1986. River loads underestimated by rat- view of approaches and techniques used in the Him- ing curves. Water Resour. Res. 22:74–76. alayan sedimentary basins. Earth Sci. Rev. 74:1–72. Gaillardet, J.; Dupre´, B.; Louvat, P.; and Alle`gre, C. J. Nyi Nyi. 1967. The physiography of Burma. Rangoon, Rangoon Arts and Science University. 1999. Global silicate weathering and CO2 consump- tion rates deduced from the chemistry of large rivers. Raymo, M. E.; Ruddiman, W. E.; and Froelich, P. N. 1988. Chem. Geol. 159:3–30. Influence of late Cenozoic mountain building on Gordon, R. 1879–1880. Report on the Irawadi River. Pt. ocean geochemical cycles. Geology 16:649–653. I–IV. Rangoon, Secretariat, 550 p. Samuelson, B. M. 1913. Note on Mr. Gordon’s discharge ———. 1885. The Irawadi River. Proc. R. Geogr. Soc. 7: scales for the Irrawaddy River at Saiktha in 1872–73 292–331. and 1875. Rangoon, Government Printing, 47 p. Humphreys, A. A., and Abbot, H. L. 1861. Report upon Socquet, A., and Pubellier, M. 2005. Cenozoic defor- the physics and hydraulics of the Mississippi River. mation in western Yunnan (China-Myanmar border). J. Asian Earth Sci. 24:495–515. U.S. Army Corps of Engineers Prof. Pap. 13. Phila- Stamp, L. D. 1940. The Irrawaddy River. Geogr. J. 95: delphia, Lippincott. 329–352. Meade, R. H. 1996. River-sediment inputs to major del- Subramanian, V., and Ittekkot ,V. 1991. Carbon transport tas. In Milliman, J., and Haq, B. U., eds. Sea level rise by the Himalayan Rivers. In Degens, E. T.; Kempe, S.; and coastal subsidence. Dordrecht, Kluwer Academic, and Richey, J. E., eds. Biogeochemistry of major world p. 63–85. rivers. Scope rep. 42–157. Chichester, Wiley. Meybeck, M.; Laroche, L.; Du¨ rr, H. H.; and Syvitski, J. Syvitski, J. P. M.; Vo¨ro¨ smarty, C. J.; Kettner, A. J.; and P. M. 2003. Global variability of daily suspended solids Green, P. 2005. Impact of humans on the flux of ter- and their fluxes in rivers. Global Planet. Change 39: restrial sediment to the global coastal ocean. Science 65–93. 308:376–380. Meybeck, M., and Ragu, A. 1995. River discharges to the Walling, D. E. 2006. Human impact on land-ocean sed- ocean: an assessment of suspended solids, major ions, iment transfer by the world’s rivers. Geomorphology and nutrients. Nairobi, United Nations Environment 79:192–216. Program, 245 p. Wilkinson, B. H. 2005. Humans as geologic agents: a deep Milliman, J. D., and Meade, R. H. 1983. Worldwide de- time perspective. Geology 33:161–164. livery of river sediments to the oceans. J. Geol. 91:1– Zeitler, P. K.; Meltzer, A. S.; Koons, P. O.; Craw, D.; Hal- 19. let, B.; Chamberlain, C. P.; Kidd, W. S. F., et al. 2001. Milliman, J. D., and Syvitski, J. P. M. 1992. Geomorphic/ Erosion, Himalayan geodynamics, and the geology of tectonic control of sediment discharge to the ocean: metamorphism. GSA Today 11:4–9.